dabetai — AI Models

Machine learning pipelines for training, evaluating, and serializing predictive models for diabetic complications.

Repository · Report Bug · Research Paper

🇬🇧 English · 🇪🇸 Español

---

About dabetai

dabetai is a comprehensive preventive ecosystem for diabetes that predicts complications like retinopathy, nephropathy, neuropathy, and diabetic foot before they become irreversible.

This repository contains the AI Models — complete pipelines for training, evaluating, and serializing machine learning models focused on predicting:

• Diabetic retinopathy
• Diabetic nephropathy
• Diabetic neuropathy
• Diabetic foot

The models are based on clinical and biometric data from the IOBP2 study and are optimized with advanced techniques such as class balancing, hyperparameter tuning, and cross-validation.

Ecosystem

Component	Repository	Stack
Mobile App	dabetai-org/mobile-app	React Native 0.79, Expo 53, Tailwind CSS
Web Portal	dabetai-org/web-app	Angular 19, Tailwind CSS
Core API	dabetai-org/api	NestJS 11, PostgreSQL, Prisma
AI Inference API	dabetai-org/ai-api	FastAPI, Python 3.11, MongoDB
AI Models (this)	dabetai-org/ai-models	Python, scikit-learn, XGBoost, PyTorch
Landing	dabetai-org/landing	Astro, Tailwind CSS

Features

• Modular Dataset Preparation — Automated per-complication data pipelines
• Multi-Algorithm Experimentation — Logistic Regression, Random Forest, LightGBM, XGBoost, SVM, AdaBoost
• Hyperparameter Optimization — Grid Search for optimal model configuration
• Model Serialization — Export trained models for production deployment
• Automatic Reporting — ROC curves, confusion matrices, feature importance visualizations

Quick Start

Prerequisites

• Python 3.11+
• pip

Setup

git clone https://github.com/dabetai-org/ai-models.git
cd ai-models
pip install -r requirements.txt

Project Structure

ai-models/
├── scripts/
│   ├── 01_prepare_datasets.py
│   ├── 02_run_experiments.py
│   └── 03_finalize_model.py
├── data/
│   ├── raw/
│   └── processed/
├── models/
├── reports/
│   └── figures/
└── requirements.txt

Usage

1. Prepare datasets

python scripts/01_prepare_datasets.py

2. Run experiments

python scripts/02_run_experiments.py

3. Finalize models

python scripts/03_finalize_model.py

Required Data

Data is based on the IOBP2 (In Control) study. Place files in data/raw/datatables/. See CITATION.md for attribution and responsible use.

Contributing

Please read CONTRIBUTING.md for branch naming, commit conventions, and PR workflow.

License

This project is licensed under the GNU General Public License v3.0 — see the LICENSE file for details.

Acknowledgments

Authors:

• Cardenas Cabal Fermín
• Ortiz Pérez Alejandro — alex03ortizperez@gmail.com
• Serrano Puertos Jorge Christian — christian.serrano.puertos@gmail.com

Advisors:

• Guarneros Nolasco Luis Rolando
• Cruz Ramos Nancy Aracely

Academic Support:

• Universidad Tecnológica del Centro de Veracruz